A7 B7 經濟部中央標準局員工消費合作社印製 i、發明説明（） 本發明係有闋霣視信號處理《路，且尤其是有鼷在所 有廣播棲準中電視倍號之鬼影信號消除電路。 回《倍號或鬼影信號是現代霣視發射中之一主要問題 。當經由空氣接收到發射信號時，可能由於高山、建築物 、和類似物髏之反射而引起鬼彩。在經由電嫌線而接收到 的«視信號中，鬼影可能由於連接器之不連鳙性而產生。 雖然鬼彩產生於無線«頻譜中，鬼影產生的程序可精確地 Μ—基本頻帶倍號中之線性失真加以棋式化。因此，将該 基本頻帶信號傳經過該鬼彩程序之線性棋式之倒反的線性 濾波器可逹成鬼彩之消除。 消除鬼影之先前技術使用例如《荷耩合裝置（CCD)橫 向濾波器的類比技術，其由於濾波器之不精確性而無法產 生顯著改進。最近*有人提出用於«波器之訓練的鬼影消 除器參考（GCR)信號。數位濾波器也己被廣泛地接受為製 作鬼影消除器之不同方法。 鬼影倍號可在主信號之前或之後抵達一接收器，因而 分別地形成光檷前鬼彩和光檷後鬼彩。消除光檷前鬼影之 濾波器是一種實際上無法實現之無限鼷衝響ii(IIR)濾波 器，其可由一棰較長的有限腯衝鬱應（FIR)濾波器加Μ近 似。光檷後鬼影可由一種IIR濾波器所消除。因此，消除 鬼影信號的典型數位濾波器是一棰FIR前嫌濾波器後面再 跟随著一組IIR回饋濾波器。 為了計算濾波器係數，從廣播站發射出一組鬼影消除 器參考（GCR)信號，一般是在一行垂直遮沒區間（VB1)内。 本紙張尺度適用中國國家標準（CNS ) A4規格（210X297公釐） ^^^1' n^i In m· (請先閱讀背面之注意事項再填寫本頁)A7 B7 Printed by the Consumers ’Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs i. Description of the invention () The present invention is based on the signal processing of 雜 霣 signal processing, and in particular there is a ghost signal cancellation circuit for TV multiples in all broadcast quasi-quasi . Back to "The double or ghost signal is one of the major problems in modern gigantic launch. When the transmitted signal is received via the air, ghost colors may be caused by reflections from mountains, buildings, and similar objects. In the «visual signal received via the electrical cable, ghosts may be generated due to the non-connectivity of the connector. Although ghost colors are generated in the wireless spectrum, the process of ghosting can be accurately chess-shaped by linear distortion in the multiple of the basic frequency band. Therefore, passing the basic frequency band signal through the linear inverse linear filter of the ghost color program can eliminate ghost color. Prior techniques for eliminating ghosting use analog techniques such as the CCD horizontal filter, which cannot produce significant improvements due to the inaccuracy of the filter. Recently * someone has proposed a GCR signal for «wave trainer training. Digital filters have also been widely accepted as different methods of making ghost removers. The ghost multiplier can reach a receiver before or after the main signal, thus forming the ghost color before and after the light filter, respectively. A filter that eliminates ghosts before light is an infinitely impulsive II (IIR) filter that is practically impossible to achieve. It can be approximated by a long finite FIR filter. The ghost after light can be eliminated by an IIR filter. Therefore, a typical digital filter for eliminating ghost signals is a FIR filter followed by a set of IIR feedback filters. In order to calculate the filter coefficients, a group of ghost canceller reference (GCR) signals are transmitted from the broadcasting station, generally within a line of vertical blanking interval (VB1). This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297mm) ^^^ 1 'n ^ i In m · (Please read the precautions on the back before filling this page)
*tT 5 經濟部中央標隼局員工消費合作社印製 A7 〜___B7_ 五、發明説明（） 濾波器的係數可由比較接收到GCR倍號與一所存檷準GCR信 號而估算出。為了消除未知的鬼影，濾波器必須能適應於 不同情況。目前的鬼影消除法是捕捉該GCR資料且將該資 料送至一分別的計算裝置，例如一組DSP或微處理器条統 ，其隨後進行計算以決定在GCR資料中之鬼彩。該糸統随 後規劃出用於視訊濂波之一组濾波器的係數。 本發明係闢於一種鬼彩消除糸统*其中在更改主動式 視訊鬼影濾波器的前嫌和回饋濾波器之係數時不需有信號 處理器或微處理器。 依據本發明，主動式視訊濾波器是使用於一視訊倍彍 之一垂直遮沒匾間Μ處理一鬼彩消除器參考（GCR)倍號和 更改該主動濾波器之係數。過濾後的鬼影參考信號與一所 存的鬼彩參考信號比較，且該差量被處理而作為更改主動 濾波器之係數的一檷誤差信號。該程序被叠代地重複直至 該誤差倍號成爲或接近零。 在一實施例中，主動濾波器的係數存在第一庫内，且一 組GCR配合濾波器（理想GCR之時域倒反）的係數被存在第二 庫内。當一般視訊倍號處理時，主動濾波器接收且過雒視 訊信號並且使用在第一庫内所存的濾波器係數移去鬼影倍 號。當一垂直遮沒區間時，接收到的鬼影參考信號被主動 濾波器反覆地過濾，且累積（平均）的過濾後鬼彩倍號與所 存的參考鬼影信號比較而使其間差量成為一誤差倍號e 。 該主動濾波器被利用在第二摩內的係數加以重新組態成爲 一組配合濾波器，並且該誤差倍號隨後被濾過Μ產生e和 本紙張尺度適用中國國家梯準（CNS ) A4規格（210X297公釐） (請先閲讀背面之注意事項再填寫本頁) 訂. 經濟部中央標準局員工消費合作社印裂 A7 B7 五、發明説明（) 理想GCR之間的數學上交互相两值，它是代表所接收到鬼 影延遲和係數之一组脈衝，dk，其中：* tT 5 Printed by the Employee Consumer Cooperative of the Central Standard Falcon Bureau of the Ministry of Economic Affairs A7 ～ ___ B7_ V. Description of the invention () The coefficient of the filter can be estimated by comparing the received GCR multiple number with a stored standard GCR signal. In order to eliminate unknown ghosts, the filter must be able to adapt to different situations. The current ghost elimination method is to capture the GCR data and send the data to a separate computing device, such as a set of DSPs or microprocessors, which then performs calculations to determine the ghost color in the GCR data. The system then plans out the coefficients of a set of filters for video sonic waves. The present invention is based on a ghost color elimination system * in which it is not necessary to have a signal processor or a microprocessor when changing the coefficients of the active video ghost filter and the coefficients of the feedback filter. According to the present invention, an active video filter is used in a video mask to process a ghost color canceler reference (GCR) multiple and change the coefficient of the active filter. The filtered ghost reference signal is compared with a stored ghost color reference signal, and the difference is processed to serve as an error signal that changes the coefficient of the active filter. This procedure is repeated iteratively until the error multiple becomes or approaches zero. In one embodiment, the coefficients of the active filter are stored in the first bank, and the coefficients of a group of GCR-matched filters (ideal GCR inverse time domain) are stored in the second bank. When processing normal video multiples, the active filter receives and passes the video signal and uses the filter coefficients stored in the first bank to remove ghost multiples. When a vertical blanking interval occurs, the received ghost reference signal is repeatedly filtered by the active filter, and the accumulated (average) filtered ghost color multiple is compared with the stored reference ghost signal to make the difference between them become a Error multiplier e. The active filter is reconfigured into a set of matched filters using the coefficients in the second motor, and the error multiple is then filtered to produce e and the paper size is applicable to China National Standard (CNS) A4 specifications ( 210X297mm) (Please read the precautions on the back before filling in this page). Order. The Ministry of Economic Affairs Central Standards Bureau employee consumer cooperatives printed A7 B7. V. Description of invention () The mathematical interaction between the ideal GCR has two values, it Is a set of pulses representing the received ghost delay and coefficient, dk, where:
N dk = X(j) *e(j+icy ] j '七 ύίΜ 在上式内x<j>是理想GCR序列（摩内係數之時域倒反序列> ，且e (j+k)是所接收到平均GCR和參考GCR之間差量。 鬼影延遲和係數（ak>被衮減（△>且加至在第一庫内之 衮減（L>濾波器係數Μ產生第一庫内之更改後濂波器係數 ，如下所示： 3·^ηβι«Γ _ (U)*3大 ί·;. ί\· • - · ；/ νΗ- 該程序製成更改係數的一種交互相_演算法則。該更 改程序繼續進行以產生交互相闋演算法則之收斂動作。 本發明和其目的及特酤將可配合附圔從下列詳細說明 和申陳專利範臞而更容易明白。N dk = X (j) * e (j + icy] j '七 ύίΜ In the above formula, x < j > is an ideal GCR sequence (inverse time domain sequence of Monet's coefficient>), and e (j + k) Is the difference between the received average GCR and the reference GCR. Ghost delays and coefficients (ak > are subtracted (△>) and added to the subtraction (L > filter coefficients M in the first library to produce the first The changed waver coefficients in the library are as follows: 3 · ^ ηβι «Γ _ (U) * 3 大 ί · ;. ί \ · •-·; / νΗ- This program makes an exchange of changing coefficients Mutual_algorithm. The modification procedure continues to produce the convergence action of the interactive phase algorithm. The present invention and its purpose and features will be more easily understood from the following detailed description and application patent specifications.
Ji , 皆·是依據本發明之一實施例的一種鬼影倍號消除条 / 統之功能方塊圏。 現在參看附_，具有鬼彩之視訊倍號施加於_入10, 經遇一组三位置開鬮12而到逹一組主動濾波器14以移除鬼 影，且接著經過一組雙位置開η 16而到達鬼彩消除視訊輪 出18。該主動濾波器14包含被一組nR回«濾波器14-2所 跟隨的一組FIR前饋濾波器14-1。該主動濾波器是由傅經 本紙張尺度適用中國國家橾準（CNS > A4規格（210X297公釐） (請先閲讀背面之注意事項再填寫本頁) --¾ -7 - 經濟部中央橾準局員工消費合作社印製 A7 B7 五、發明説明（） 過多工器22而到連主動濾波器的A庫20内所存係數組態而 成。第二係數庫24儲存著將主動濾波器14組態成為一組配 合濾波器的係數，將在下面說明。 在垂直遮沒區間時，濾波器14被組態為一組配合濾波 器且所接收到的GCR被送經該濾波器Μ決定GCR的極性和時 序。而且，接收到的GCR經由開闋37被存入FIFO 36。在 FIFO 36内的GCRfil後傳經主動濾波器14,其被组態為一種 初姶未收斂的鬼彩消除濾波器，再經由開關26,互斥或閘 28,到逹一組包含加法器30和FIFO 32的累稹或平均回路 。該互斥或閘28反應來自正/負GCR檢拥及時序電路34之 信號，電路34在主動濾波器被以來自第二庫24的係數组態 成為一組配合濾波器時決定進入的GCR信號是正或負。同 時，GCR信號被存在FIFO 36内Μ便有充裕時間傳經濾波器 14。檢測器34也量測存在FIFO 36内GCR信號的捕捉時序Μ 控制FIFO信號的放出且移去圖素失序抖動效應。 起初，存在摩20内的係數將主動濾波器14組態成為一 種全通濾波器*但是當重算的係數收斂時，該等係數將主 動濾波器14組態成為GCR信號的一種鬼影消除濾波器。至 少八値圈場（或八的倍數痼〉被平均器30和FIFO 32捕捉到 且接著被加法器電路30相加或相減（若需要的話）並且累稹 於FIFO 32。在FIFO 32内之累積值或平均GCR與存在FIFO 40内之參考GCR倍號一起被送至加法器38。平均GCR和參考 GCR在加法器38被相減Μ產生誤差倍號e 。誤差信號e是 乾淨GCR倍號和所接收到且被遇濾的GCR信號之間的差量。 本紙張尺度適用中國國家標準（CNS ) A4規格（210X 297公釐） (請先閱讀背面之注意事項再填寫本頁) 装_ 訂. 8 經濟部中央標準局員工消費合作社印製 Α7 Β7 五、發明説明（） 該誤差倍號接蓿被傅經過開關12而回至主動濾波器14,其 現在被庫24規劃成爲一組配合濾波器，該配合濾波器產生 代表所接收到鬼彩延遲和係數的一組脲衝dk。該配合濾波 器產生一組數學交互相鼷值且產生dk值如下： dk = Tix^)'e^+k^!\ 其中x(j)是理想GCR序列（庫中係數之時域倒反序列），且e (j+U是平均GCR和參考GCR之間差量。 該dk佶號接著被一組△移位器42加Μ衮減且被加法器 44加至存在摩20内之係數，該等係數傅經過在46所示之一 組衮減移位器，其將該等係數乘M (1-L)。加法器44的綸 出依據下列方程式成為新的更改係數而存在庫20内： !> ： i 丨Ή t__—I—* 庫20内的新係數接著用M將主動濾波器14重新規劃成為一 組鬼彩消除濾波器。利用在庫20内的新係數，取平均值的 處理再次地開始於垂直遮沒區間以搜集新的過濾GCR信號 。一組新的差量信號（e〉如上所述地再被處理以得到一組 新的dk佶號，且該組新的dk信號與舊的ak值組合在一起。 逭樣會產生令dk等於或趨近零的轉變動作。或者是，在庳 24内的配合濾波器係數可被載Μ—組平均接收鬼彩GCR信 號。造裂成最小平方和均值演算法則。 本紙張尺度適用中國國家標隼（CNS ) Α4規格（210><297公釐〉 (請先閱讀背面之注意事項再填寫本頁) 裝- 訂. 9 五、發明説明（ A7 B7 經濟部中央標準局員工消費合作社印製 當轉至具有鬼彩的新頻道時，可Μ進行一種初始加速 收斂。在取得初始8鹤，或8的倍數値圏場平均值，並且 存入FIFO 32之後，只利用在初始平均值内之資料，可以 進行多次處理程序W加速初始的收斂。當進行初始加速收 斂的處理時，具有鬼彩的視訊将被傅經過MAC 14經一圓場 的時間。 接收到的鬼影GCR之初始平均值被搜集且保留在FIFO 32内。由於隨後的步《將改寫FIFO 32, FIFO 32的内容將 經由開《137而複製至FIFO 36。在造被製操作時，沒有參 考GCR之相減操作發生於加法器38。使用初始A庫20之係 數ak將FIFO 36之内容放出經過MAC 14且其結果一次地被 存於FIFO 32内，而不餺有平均之操作。 在FIFO 32和參考GCR 40内的資料在38處被相減以產 生誤差信號e 。逭誤差信號e被送至設定成具有庫24之係 數的一組配合濾波器之MAC 14,因而產生一組dk信號，其 經過移位器46，44、42以更改在庫20内之ak值。逭些在庫 20内之新的ak係數被用以規劃MAC 14,接著存在FIFO 36 內之初始搜集平均GCR經由MAC 14放出，其結果未被平均 而存入FIFO 32内。在一麵場之時間内，該程序被重嫌數 次（當該時間内，作用視訊被旁通，加鬼彩）。當該時間用 完後，現行視訊被切換經過MAC 14,其現在被Μ庫20內之 最新ak係數加Μ規劃。一組新的平均GCR，使用以最新ale 鬼影消除係數而設定的濾波器14加Μ遇濾的GCR，被以先 前描述方法搜集以更改庫20之係數（初始平均GCR将失去且 (請先閱讀背面之注意事項再填寫本頁) 裝· 訂· 本紙張尺度適用中國國家標準（CNS ) A4規格（210 X 297公釐） 10 A7 B7 經濟部中央標準局貝工消費合作杜印製 五、發明説明（） 被重寫，但它已不再被使用>。 收斂率可被改變。其一方法是董度dk倍號的能量，且 若dk是處於高倍》準位則該信號並未接近收斂。一組檢拥 器準位量度«路50被用Μ改變分別在移位器42和移位器46 内的值△和L。快速收斂率用於高dk值（且在頻道中之第一 次調整>*而較低收斂率用於接近收斂時。而且，準位量 度«路50也可加上«路以檢測dk準位中之突然增加。«後 可決定是否應用新的dk值或將該新值與下一 dk值比較而明 白原先增加值是真正值或只是一種暫態值。逭提供避免使 用暫態不良dk值不正確地更改在鬼彩消除条統中過濂係數 之方法。而且，當接近收斂時，該更改可被凍结Μ停住係 數之隨機小量變化。埴將減少因小量隨機變化所引起在霣 視幕上消除鬼彩後視訊之閃動可能性。使用比凍结時較高 的解凍準位將形成一棰磁滯迺路，可防止當準位在凍結臨 限值附近時引起的抖動現象。 爲了使《路34能決定正性或爽性的GCR信號，濾波器 14被Μ來自庫24之係數設定成爲一捶配合濾波器棋式。Μ 所接收到的GCR_入而言，濾波器14的綸出看起來像一組 大腯衝（代表主要的GCR倍號〉以及一些較短脈衝（代表鬼 彩）。正負檢測器34決定圈素時序以及脈衡為正或負。逭 接著用Μ經過邏輯闢28而控制平均加法器30。檢澜器霣路 34可以是記錄最大正值或負值的一種臨限檢测器，且該最 大络對值被S擇以決定正性或負性GCR值。而且，該脲衝 之時序也可舆其預期位置比較以決定是否因抖動的同步分 本紙張尺度適用中國國家標準（CNS ) Α4規格（210X297公釐） ---------/ 装-- (請先閲讀背面之注意事項再填寫本頁) 丁♦. -'3 11 經濟部中央標準局員工消費合作社印製 A7 B7 i、發明説明（） 離器54而有圔索失去時序，並且逭資訊可送至FIFO和多工 控制器56以調整FIFO 36之時序而避免在平均FIFO 32内圔 素不配合情形。失去時序將使条統作用如同有額外鬼影且 同步和陣列將無法被適當地平均，導致誤差信號（e>和dk 信號具有額外的混亂。逋将擾亂在庫20内之濾波器係數且 將引起主動濾波器14失去規割及新鬼影出現在視訊屏幕上 ，或甚至使濾波器振盪或不穩定。 主動濾波器14包含一組前饋PIR濾波器14-1，一組回 餱IIR濾波器14-2, 一組加法器14-3以及一組雙位置開關 14-4。在正常鬼影消除棋式中且當在垂直區間進行計算時 ，利用來自庫20之係數，該主動濾波器14被設定成一種 FIR和IIR濾波器構造。這可由於設定開鼷14-4將濾波器 14-2的輪出連至加法器14-3而使濾波器14-2作用如一組 IIR。加法器14-3的鑰出是鬼影消除棋式輪出。作爲一組 配合濾波器時，開關14-4將加法器14-3連至接地。開鼷 14-4可當作及閘使用，高準位控制線可通過倍號且使濾波 器14-2成為一組IIR，低準位產生零值且使濾波器14-2成 爲一組FIR。該配合濾波器輪出係取自濾波器14-2的輸出 ，當它用以檢測GCR時序和極性時被饋至正/貪檢測器34 ，且當它用於係數更改時提供dk倍號綸出至△移位器42和 準位檢測器50。因此，當使用在庫20内之係數時，利用切 換開關14-4M使加法器14-3能加上濾波器14-2之綸出，使 得主動濾波器14被組態成爲鬼影消除棋式之一捶FIR和 IIR。加法器14-3的_出足經由平均器30謓至FIFO 32且經 本紙張尺度適用中國國家樣準（CNS > A4規格（210X297公釐〉 ---------f 裝—— (請先閲讀背面之注意事項再填寫本頁) 訂- -12 - 經濟部中央標準局員工消費合作社印製 A7 _B7 __ 五、發明説明（） 由開两16嫌至糸統轅出18的一组輪出。欲轉換成一組配合 濾波器時，將開P14-4切換至接地而使IIR動作失去作用 且串列地建接濾波器14-1和14-2,主動濾波器即被重新組 態成為一組大的FIR濾波器。當該主動濾波器被組態成一 組鬼影消除濾波器時，存在庫20內的係數供應至濾波器 14-1和14-2。 當主動濾波器14被用Μ進行GCR處理時，在垂直區間 之一部份中，作用性的視訊綸入在濾波器附近被錯開而至 輪出，或者是在垂直匾間中，從產生器58產生被替代的垂 直波形。 所示的演算法依據IIR濾波器的輪出而更改濾波器係 數。如所習知，本演算法可能由於在回嫌係數的更改中之 固有之非線性而不穩定。本架構可容易地被延伸以纳入更 穩定演算法，例如在相同指定人之1993年5月7日建檑待 決專利案號08/ 060,374中所述。所示演算法使用美國GCR 作為一例。加Μ修改後，本架構也可用於日本GCR和其他 GCR 〇 上面已說明一棰鬼影消除条統，其中該主動濾波器被 連續調整以消除鬼影而不需信號處理器或撤處理器。雖然 本發明已參考一特定實施例加Μ說明，該說明是本發明之 一種展示而非作為本發明之限制。對於熟習本技術者可知 本發明可有各棰修改和應用而不脫離所附申請專利範圍所 定義之本發明的真正精神和範疇。 本紙張尺度適用中國國家標準（CNS ) Α4規格（210Χ297公釐） (請先閱讀背面之注意事項再填寫本頁) 装· 訂- -13 -Ji, Jie · is a function block of a ghost double number elimination bar / system according to an embodiment of the present invention. Now refer to the attached _, video multiplier with ghost color is applied to _10, after encountering a set of three-position switch 12 to a set of active filter 14 to remove ghosts, and then through a set of two-position switch η 16 and reached the ghost color elimination video round 18. The active filter 14 includes a set of FIR feedforward filters 14-1 followed by a set of nR feedback filters 14-2. The active filter is made by Fu Jingben. The paper standards are applicable to the Chinese National Standard (CNS & A4 specifications (210X297mm) (please read the precautions on the back before filling out this page) --¾ -7-Central Ministry of Economic Affairs A7 B7 printed by the Bureau ’s Consumer Cooperative. V. Description of the invention () The configuration of the coefficients stored in the A library 20 connected to the active filter by the multi-tool 22. The second coefficient library 24 stores the configuration of the active filter 14 The coefficients that become a set of matched filters will be described below. In the vertical blanking interval, the filter 14 is configured as a set of matched filters and the received GCR is sent through the filter M to determine the polarity of the GCR And the timing. Moreover, the received GCR is stored in the FIFO 36 via the gate 37. The GCRfil in the FIFO 36 is passed through the active filter 14, which is configured as a ghost color elimination filter that has not yet converged, and then Via switch 26, mutex or gate 28, a set of accumulated or average loops including adder 30 and FIFO 32. This mutex or gate 28 reflects the signal from the positive / negative GCR detection and timing circuit 34, the circuit 34 The active filter is replaced by the coefficients from the second bank 24 When the state becomes a set of matched filters, the GCR signal that is entered is determined to be positive or negative. At the same time, the GCR signal is stored in the FIFO 36 and M has sufficient time to pass through the filter 14. The detector 34 also measures the capture of the GCR signal stored in the FIFO 36 Timing M controls the release of the FIFO signal and removes the pixel-out-of-sequence jitter effect. Initially, the coefficients in the motor 20 configure the active filter 14 as an all-pass filter * but when the recalculated coefficients converge, the coefficients The active filter 14 is configured as a ghost elimination filter for GCR signals. At least eight circle fields (or multiples of eight) are captured by the averager 30 and the FIFO 32 and then added or added by the adder circuit 30 Subtract (if necessary) and accumulate in FIFO 32. The accumulated value or average GCR in FIFO 32 is sent to adder 38 together with the reference GCR multiple number stored in FIFO 40. The average GCR and reference GCR are in adder 38 The subtracted Μ produces an error multiple e. The error signal e is the difference between the clean GCR multiple and the received and filtered GCR signal. This paper scale applies the Chinese National Standard (CNS) A4 specification (210X 297 Mm) (please Read the precautions on the back first and then fill out this page.) _ _ Order. 8 Printed Α7 Β7 by the Employee Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs Ⅴ. Invention Description () This error multiple is connected to the active filter by Fu after passing switch 12 14, which is now planned by library 24 as a set of cooperating filters that produce a set of urea pulses dk representing the received ghost delay and coefficients. The cooperating filters produce a set of mathematically interactive phase values and The resulting dk values are as follows: dk = Tix ^) 'e ^ + k ^! \ Where x (j) is the ideal GCR sequence (inverse time-domain sequence of coefficients in the library), and e (j + U is the average GCR and reference Difference between GCR. The dk number is then added by a set of delta shifters 42 and subtracted by an adder 44 to the coefficients present in Mo 20. These coefficients pass through a set of shift subtractors shown in 46, which Multiply these coefficients by M (1-L). The output of the adder 44 becomes a new modified coefficient according to the following equation and is stored in the library 20:! ≫: i 丨 Ή t __— I— * The new coefficient in the library 20 is then used to re-plan the active filter 14 into a Group ghost color elimination filter. Using the new coefficients in the library 20, the averaging process starts again in the vertical blanking interval to collect new filtered GCR signals. A new set of difference signals (e> is processed as described above to obtain a new set of dk numbers, and the new set of dk signals is combined with the old ak value. A sample will produce dk equal to Or approaching zero transition action. Or, the matching filter coefficients in Bo 24 can be loaded with M-groups to receive the GCR signal on average. Split into the least squares and mean algorithm. The paper scale is applicable to the Chinese national standard Falcon (CNS) Α4 specification (210 > < 297mm> (Please read the precautions on the back before filling out this page) Binding-Order. 9 V. Description of invention (printed by A7 B7 Employee Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs When switching to a new channel with ghost colors, an initial accelerated convergence can be performed. After the initial average value of 8 cranes, or multiples of 8 is obtained and stored in the FIFO 32, only the initial average value is used Data can be processed multiple times to accelerate the initial convergence. When the initial accelerated convergence process is performed, the video with ghost colors will be passed through the MAC 14 for a round field time. The initial average value of the received ghost GCR Collected and secured In the FIFO 32. Since the subsequent step "will rewrite the FIFO 32, the content of the FIFO 32 will be copied to the FIFO 36 via the opening" 137. In the fabricating operation, the subtraction operation without reference to the GCR occurs in the adder 38. The coefficient ak of the initial A library 20 is used to release the content of the FIFO 36 through the MAC 14 and the result is once stored in the FIFO 32 without averaging. The data in the FIFO 32 and the reference GCR 40 are at 38 It is subtracted to produce the error signal e. The error signal e is sent to a set of MAC 14 with a set of matched filters having the coefficients of the bank 24, thus generating a set of dk signals, which pass through the shifters 46, 44, 42 To change the ak value in the library 20. Some new ak coefficients in the library 20 are used to plan MAC 14, and then the initial collected average GCR stored in the FIFO 36 is released through the MAC 14, and the result is not averaged and stored in the FIFO Within 32. Within a field time, the program was revisited several times (when this time, the function video was bypassed and added ghost color). When this time was used up, the current video was switched through MAC 14, It is now planned by the latest ak coefficient in M library 20 plus M. A new set of The average GCR, using the filter 14 set with the latest ale ghost elimination coefficient and the GCR filtered, is collected by the method described previously to change the coefficient of the library 20 (the initial average GCR will be lost and (please read the note on the back first Please fill in this page again for details) Binding · Order · This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 10 A7 B7 Printed by the Cooperative Consumer Cooperation of the Central Standards Bureau of the Ministry of Economy V. Invention description () Was rewritten, but it is no longer used>. The convergence rate can be changed. One method is the energy of Dongdu's dk times, and if dk is at a high power level, the signal is not close to convergence. A set of detector level measurements «way 50 is changed by M to the values △ and L in the shifter 42 and shifter 46, respectively. The fast convergence rate is used for high dk values (and the first adjustment in the channel > * while the lower convergence rate is used for near convergence. Moreover, the level measurement «Route 50 can also be added with« Route to detect the dk level Sudden increase in the bit. Afterwards, you can decide whether to apply the new dk value or compare the new value with the next dk value to understand whether the original increase value is a real value or just a transient value. The Japanese provides to avoid the use of transient dk The value incorrectly changes the method of overlying the coefficient in the ghost color elimination system. Moreover, when close to convergence, the change can be frozen by random small changes in the coefficient of anchoring. It will reduce the number of random changes due to small random changes. Causes the possibility of eliminating the flashing of the ghost video after the ghost screen. Using a higher thaw level than when freezing will form a hysteresis loop, which can prevent the level from being caused when the freezing threshold is near The jitter phenomenon. In order to make the channel 34 determine the positive or cool GCR signal, the filter 14 is set by the coefficient from the library 24 as a hammer match filter chess. In terms of the received GCR_ input , The filter 14's lun out looks like a set of big churn (represent The required GCR multiplier number> and some shorter pulses (representing ghost colors). The positive and negative detectors 34 determine the cycle timing and pulse balance to be positive or negative. Then use M to control the average adder 30 through logic 28. The device path 34 can be a threshold detector that records the maximum positive or negative value, and the maximum complex pair value is selected by S to determine the positive or negative GCR value. Moreover, the timing of the urea pulse can also be The expected position is compared to determine whether the paper size of the synchronous sub-sheet due to jitter is applicable to the Chinese National Standard (CNS) Α4 specification (210X297 mm) --------- / Packing-(Please read the precautions on the back first (Fill in this page again) Ding. -'3 11 A7 B7 printed by the employee consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs i. Description of invention () The device 54 loses the timing due to the device, and the information can be sent to FIFO and multiplex The controller 56 adjusts the timing of the FIFO 36 to avoid uncoordinated situations in the average FIFO 32. Losing the timing will make the system function as if there are extra ghosts and the synchronization and array will not be properly averaged, resulting in an error signal (e > And dk signals have additional confusion. The filter coefficients in the library 20 will cause the active filter 14 to lose its rule and new ghosts appear on the video screen, or even make the filter oscillate or unstable. The active filter 14 includes a set of feed-forward PIR filters 14-1, a set of lingering IIR filters 14-2, a set of adders 14-3 and a set of two-position switches 14-4. In the normal ghost elimination chess pattern and when calculating in the vertical interval, use From the coefficients of the library 20, the active filter 14 is set to a FIR and IIR filter configuration. This can be filtered by setting the switch 14-4 to connect the round of the filter 14-2 to the adder 14-3 The device 14-2 functions as a set of IIRs. The key out of the adder 14-3 is a ghost elimination chess round. As a set of matched filters, switch 14-4 connects adder 14-3 to ground. Open reed 14-4 can be used as a brake, the high level control line can pass through the multiplier and make the filter 14-2 become a set of IIR, the low level produces a zero value and makes the filter 14-2 become a set of FIR . The matching filter round is taken from the output of the filter 14-2, which is fed to the positive / corrupt detector 34 when it is used to detect the timing and polarity of the GCR, and provides dk times the number of fibers when it is used to change the coefficient Out to the delta shifter 42 and the level detector 50. Therefore, when the coefficients in the library 20 are used, the switch 14-4M is used to enable the adder 14-3 to add the filter 14-2, so that the active filter 14 is configured into a ghost elimination chess style One beat FIR and IIR. The adder 14-3's output is passed to the FIFO 32 through the averager 30 °, and the Chinese national standard (CNS & A4 specifications (210X297 mm) is applied to this paper standard---------- f equipment-( Please read the precautions on the back and then fill out this page) Order--12-Printed by the Ministry of Economic Affairs Bureau of Central Standards Employee Consumer Cooperative A7 _B7 __ Fifth, the description of invention () from open two 16 suspect to Shi Tongyuan out of a group of 18 Turn out. When you want to convert to a set of matched filters, switch P14-4 to ground to make the IIR action invalid and connect the filters 14-1 and 14-2 in series, the active filter is reconfigured Becomes a large set of FIR filters. When the active filter is configured as a set of ghost elimination filters, the coefficients stored in the library 20 are supplied to the filters 14-1 and 14-2. When the active filter 14 is When performing GCR processing with Μ, in a part of the vertical interval, the active video input is staggered near the filter to the round out, or in the vertical plaque, the replaced vertical is generated from the generator 58 Waveform. The algorithm shown changes the filter coefficients based on the round of the IIR filter. As is known, The algorithm may be unstable due to the inherent non-linearity in the change of the suspicion coefficient. This framework can be easily extended to include a more stable algorithm, for example, on May 7, 1993 for the same designator. As described in Patent Case No. 08 / 060,374. The algorithm shown uses the US GCR as an example. After modification, this architecture can also be used for Japanese GCR and other GCRs. A ghost ghost elimination rule has been described above, in which the active The filter is continuously adjusted to eliminate ghosting without the need for a signal processor or a de-processor. Although the present invention has been described with reference to a specific embodiment, the description is a display of the present invention and not a limitation of the present invention. Those skilled in the art can know that the invention can be modified and applied without departing from the true spirit and scope of the invention as defined in the scope of the attached patent application. This paper scale is applicable to the Chinese National Standard (CNS) Α4 specification (210Χ297 mm) (Please read the precautions on the back before filling in this page) Binding · Order--13-